huihui-ai/Huihui-gemma-4-31B-it-abliterated-v2
Abliterated vs google/gemma-4-31B-it. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 55 of 60 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/4, Norms 0/523, other 0/62, MLP 55/261, Attention 55/338
6 derivatives checked
- mradermacher/Huihui-gemma-4-31B-it-abliterated-v2-i1-GGUF: Quantization (fewer layers)
- mradermacher/Huihui-gemma-4-31B-it-abliterated-v2-GGUF: Quantization (fewer layers)
- groxaxo/Huihui-gemma-4-31B-it-abliterated-v2-GGUF: Quantization (fewer layers)
- lyf/Huihui-gemma-4-31B-it-abliterated-v2-NVFP4: Quantization
- lazywoof/Huihui-gemma-4-31B-it-abliterated-v2-Q6_K-GGUF: Quantization (fewer layers)
- win10/K1-31B-v5: LoRA-style
6 other repos hold byte-for-byte identical weight files.